The project aims at cancellation of white noise from a speech signal adaptively using adaptive filters that change their weights based on the LMS Algorithm.
A simulink model has been provided along with the real time simulation video.
2. BITS Pilani, Pilani Campus
ANC SPEECH SIMULINK
A Simulink Model that adaptively filters out the noise in speech
Speech signal that runs indefinitely during the simulation s(n)
WGN n1(n) is generated by the simulink block and is filtered using a FIR Filter
The filtered noise n2(n) is band limited and is now added to the Speech Signal
The “signal + noise” is our desired signal d(n) = s(n) + n2(n)
The LMS Block takes an input as a reference noise n1(n) for the purpose of
adaptive filtering
The output e(n) is played via the audio device
4. BITS Pilani, Pilani Campus
PARAMETERS AND VALUES
Input signal s(n) - represented by an .aiff file containing speech sampled at 15 KHz
Source noise signal n1(n) - White Gaussian Noise with 0 mean and variance = 0.1 (20dB)
Direct-form FIR Low pass filter 10th order. - Used to create a second noise n2(n) that is
correlated with the source noise n1(n)
For this simulation, the Adaptive Filter length is 32.
The step-size parameter μ was chosen to be 0.0008.
11. BITS Pilani, Pilani Campus
Colored Noise : Selected Group of Frequencies are affected more by it.
Examples : A Jet plane’s Engine Noise
White Gaussian Noise can be replaced by a colored noise source.
Colored Noises can be recorded manually using a Condenser Microphone
We saw ANC in real time when the Speech signal was corrupted by White
Gaussian Noise
Next, ANC SPEECH SIMULINK will be simulated for various colored noises and
comparisons will be provided
FUTURE PLAN OF ACTION